There is a chicken/egg problem with restarting SolrCloud pods.

• We need to be able to connect to a pod in order to do any necessary recovery of replicas contained on that Solr Node. Therefore the Service for that node needs to point to the pod.
• We need to ensure that a rolling restart of the StatefulSet does not start on the next pod until all replicas are healthy in the pod. The only way we can do that through the provided statefulSet mechanisms is through a robust readinessCheck.

So if we want both of these to be true, then the SolrNode services (both individual node services, and the headless service) need to route to pods regardless of whether that node is ready or not. This can be achieved by using the following option:

Service:
  Spec:
    PublishNotReadyAddresses: true